Speech-to-Text · Multilingual

Arabic

Language

Request a new language

Tell us which language you want added to this benchmark.

Benchmarks

X
Y

CER

Character error rate — word scoring mismeasures this language.

MunsitMunsitTry →2.7%
OpenAIGPT-4o TranscribeTry →3.3%
AssemblyAIAssemblyAIUniversal-3.5 ProTry →3.5%
MetaMetaMuse Voice Transcribe 1.0Try →3.6%
Sonioxstt-rt-v5Try →3.6%
NariNariQwen3-ASRTry →3.9%
AlibabaQwen3-ASR Realtime4.1%
ElevenLabsScribe v2 RealtimeTry →4.2%
xAIGrok STTTry →4.7%
OpenAIGPT-4o-mini TranscribeTry →4.9%
MunsitMunsiten-arTry →4.9%
HamsaHamsaS3Try →5.0%
ModulateModulateVelma 2 multilingualTry →5.0%
DeepgramNova-3Try →7.6%
CartesiaInk-WhisperTry →10.1%
0.0%5.0%10.0%15.0%

Finalize latency

Region

Time from the caller stopping to the finalized transcript, not the first partial — the agent cannot reply until it lands.

DeepgramNova-3Try →317ms · 385ms
Sonioxstt-rt-v5Try →520ms · 745ms
OpenAIGPT-4o-mini TranscribeTry →908ms · 1.14s
AlibabaQwen3-ASR Realtime924ms · 1.07s
MunsitMunsitTry →987ms · 1.09s
OpenAIGPT-4o TranscribeTry →1.05s · 1.36s
AssemblyAIAssemblyAIUniversal-3.5 ProTry →1.49s · 1.82s
ElevenLabsScribe v2 RealtimeTry →1.99s · 2.13s
0ms300ms600ms900ms1.20s

p50 p90

Time to first token

Region

How long before the first word of the transcript arrives.

MunsitMunsitTry →1.06s · 1.72s
AssemblyAIAssemblyAIUniversal-3.5 ProTry →1.07s · 1.78s
DeepgramNova-3Try →1.25s · 2.26s
Sonioxstt-rt-v5Try →1.51s · 2.21s
ElevenLabsScribe v2 RealtimeTry →2.18s · 2.28s
AlibabaQwen3-ASR Realtime6.74s · 7.62s
OpenAIGPT-4o TranscribeTry →7.76s · 13.3s
OpenAIGPT-4o-mini TranscribeTry →7.79s · 13.4s
0ms2.00s4.00s6.00s8.00s

p50 p90

Cost

lower is better

$ per 1,000 minutes of audio · vendor list price, the same in every language